def calculate_perplexity(transactions: List[List[int]], predictor: Predictor) -> float:
perplexities = []
for tr in transactions:
out = predictor.predict_json(
{
"transactions": tr,
"amounts": tr
}
)
perp = math.exp(out["loss"])
perplexities.append(perp)
return sum(perplexities) / len(perplexities)
def _caching_prediction(model: Predictor, data: str) -> JsonDict:
"""
Just a wrapper around ``model.predict_json`` that allows us to use a cache decorator.
"""
return model.predict_json(json.loads(data))
def predict_json(self, input_json: Dict[str, Any], predictor: Predictor):
input_json = input_json if input_json else {
"sentence": "A good movie!"
}
output = predictor.predict_json(input_json)
return output
def predict_json(input_json: Dict[str, Any], predictor: Predictor):
input_json = input_json if input_json else {"sentence": "A good movie!"}
output = predictor.predict_json(input_json)
print(output)
